Refactor getClosestNeighbour

[jtaveau]

I took the initiative to refactor this function as it's the most expensive one of the project (computationnally speaking).
The results looked great on hotspot, but I had only a small bag of 15s to test it, would you have a bigger one?

Before refactor:

After refactor:

[nachovizzo]

@jtaveau thanks a lot for the initiative. Indeed the hot path from the code is looking for nearest neighbors. As it has been the case for most systems out there as well

I'll review the code and test this intensively, with all the datasets we know were it's working and come back to you. It might take some time only

PS: thanks for fixing the typos 😂

[nachovizzo]

Looks good to me! I tried on a few datasets and it's working, although I do not see any real improvement in terms of speed.

I think I prefer the other implementation since the 3 tasks that the lambda function is doing where defined in 3 different steps. With this implementation they all tree tasks got merged into quite a few nested loops, which complicates redability for new users

I will let Tiziano and Benedikt comment and test on this particular PR

Great job @jtaveau

[nachovizzo]

Closing this to avoid stalling, feel free to re-open when you manage to benchmark it!

Best!